CN109738910A - A kind of curb detection method based on three-dimensional laser radar - Google Patents
A kind of curb detection method based on three-dimensional laser radar Download PDFInfo
- Publication number
- CN109738910A CN109738910A CN201910077748.4A CN201910077748A CN109738910A CN 109738910 A CN109738910 A CN 109738910A CN 201910077748 A CN201910077748 A CN 201910077748A CN 109738910 A CN109738910 A CN 109738910A
- Authority
- CN
- China
- Prior art keywords
- curb
- laser radar
- point
- scanning
- density
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 239000000284 extract Substances 0.000 claims abstract description 7
- 238000005070 sampling Methods 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000005855 radiation Effects 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000011160 research Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 7
- 230000008901 benefit Effects 0.000 description 7
- 241000196324 Embryophyta Species 0.000 description 6
- 238000005286 illumination Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000007689 inspection Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000006641 stabilisation Effects 0.000 description 2
- 238000011105 stabilization Methods 0.000 description 2
- 241000221931 Hypomyces rosellus Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention belongs to traffic road circumstances in intelligent automobile technology to perceive field, it is related to a kind of curb detection method based on three-dimensional laser radar, the point cloud data that surrounding road environment is obtained using the three-dimensional laser radar being installed on vehicle, is partitioned into ground using stochastical sampling consistency algorithm；Wider threshold value is arranged according to a variety of geometrical characteristics of curb to be judged, using neighborhood relationships between each scanning element of same scanning slice, extracts curb candidate point；According to curb data point in density and road direction continuous feature, curb candidate point is clustered using density-based algorithms and removes the larger and isolated noise spot of density；Qualified curb candidate point is fitted with weighted least-squares method finally, improves fitting accuracy.The present invention comprehensively considers curb continuous feature in density and road direction and is denoised according to multiple feature extraction curb candidate points of curb, so that final detection error is small, precision is high.
Description
Technical field
The invention belongs to traffic road circumstances in intelligent automobile technology to perceive field, be related to a kind of based on three-dimensional laser radar
Curb detection method.
Background technique
As the economic sustained and rapid development in China and social are constantly progressive, people's lives level is significantly improved, China
Automobile industry rapid development, automobile has become the essential vehicles of people's daily life, but automobile industry
Fast development also brings the major issues such as traffic safety.Such issues that in order to solve, intelligent auxiliary driving system for automobiles are met the tendency of
And it gives birth to.Intelligent auxiliary driving system for automobiles refers to that car-mounted computer detects surrounding traffic environment letter by road environment sensory perceptual system
Breath proposes that drive advice or auxiliary driver carry out driver behavior by processing for driver, to reduce traffic accident
Rate.For many intelligent auxiliary driving system for automobiles, the key realized is the detection and identification to road ahead curb.The skill
Art can greatly reduce traffic accident, reduce personnel's property loss.
Detection method can be divided into two classes at present: first is that the detection method based on machine vision；Second is that being based on laser radar
Detection method.
Detection algorithm based on machine vision is all the preferred research direction of researcher all the time, is because of phase equipment
There is price is low to be easy to industrialization, small in size to be easily installed, the advantages that processing speed is fast.But this technology vulnerable to weather, illumination,
The interference that the environment such as shade influence, so that detection accuracy is low.Laser radar is the important biography that intelligent vehicle obtains external environmental information
One of sensor, it is received from the reflected information of target by radar receiver to objective emission laser, passes through comparison transmitting letter
Difference number between reflection signal, and carries out signal processing appropriate, to obtain the related data of target, as target away from
From, three-dimensional coordinate, reflected intensity, return to layer etc..Laser radar have do not influenced by factors such as weather, illumination, to shade noise
Insensitive equal good characteristics.And laser radar has the advantages that scanning distance is remote, investigative range is wide, detection accuracy is high.Therefore
Road environment is perceived using laser radar with preferable robustness and rapidity, is had on intelligent automobile is unmanned preferably
Application prospect.
Also occur carrying out the technology of curb detection, such as Chinese patent using laser radar in existing technology
CN104636763A uses four line laser radars as range sensor, with the equal spy of the slope of same scanning slice curb point
Sign carries out threshold decision to the slope between curb point, and application is based on the improved COBWEB algorithm of Euclidean distance to curb number
Clustering is carried out according to collection, is finally fitted curb with least square method.But four line laser radar scanning lines are less, and cannot
Three-dimensional scanning circumstance is provided, detects available environmental data more abundant using three-dimensional laser radar, is conducive to automatic
The application of driving technology.Such as Chinese patent CN107272019A carries out gradient to the point in scan line using three-dimensional laser radar
Filtering obtains curb candidate point, and carries out conic fitting to qualified curb candidate point, obtains curb testing result.
But this method only considers that Gradient Features extract curb candidate point, and detection accuracy is not high.
Summary of the invention
Based on problem of the existing technology, the invention proposes a kind of curb detection side based on three-dimensional laser radar
Method constructs a variety of curb geometrical characteristics according to the neighborhood relationships between scanning element, carries out threshold decision, guarantees every kind of feature not
Meeting missing inspection, comprehensively considers curb continuous feature in density and road direction and is denoised, and has detection accuracy higher, detection
The small advantage of error, and curb can be detected under different illumination conditions, it is not illuminated by the light condition influence, stabilization accurately detects
Road edge.
A kind of curb detection method based on three-dimensional laser radar of the invention, using including following scheme, such as Fig. 1 institute
Show, comprising the following steps:
S1, road environment is scanned by three-dimensional laser radar, obtains original point cloud data, and with three-dimensional laser thunder
Up to the three-dimensional system of coordinate for establishing out detection vehicle for origin；
S2, ground segmentation is carried out using stochastical sampling consistency algorithm to the original point cloud data, extracts road surface
Point cloud data；
S3, by the point cloud data on the road surface according to curb data point feature, carry out threshold decision, extract curb candidate
Point；
S4, according to the density feature of the curb candidate point, the curb candidate point of each scanning slice is used and is based on density
Algorithm DBSCAN carry out cluster denoising；
S5, according on road direction have successional feature, will cluster denoising after have successional curb candidate point
It is clustered again；
S6, least square method fitting is weighted to the curb candidate point after clustering again；Obtain the detection knot of road edge
Fruit.
Further, in step S1, laser radar is horizontally arranged at vehicle roof, using laser radar center as origin o,
X, y, z axis passes through origin o and the orthogonal radar fix system for establishing detection vehicle, and y-axis is parallel to the ground and inspection is directed toward in direction
The headstock of measuring car, x-axis is parallel to the ground and the right-hand direction of vehicle headstock is directed toward in direction, and z-axis faces upward perpendicular to ground.
Wherein, the point being reflected back by the laser scanning that three-dimensional laser radar emits to body surface, is used to form original
Beginning point cloud data；Original point cloud data includes three-dimensional coordinate, reflected intensity, scanning slice, distance and the azimuth of multiple objects.
Further, the original point cloud data that obtains in S1 carries out ground segmentation, due to three-dimensional laser radar it is angled to
On scan line, also angled downward scan line only considers the laser radar scanning of angle downward radiation when detecting curb
Line, and according to the relationship of the detection accuracy of radar and distance, research regional scope；Using stochastical sampling consistency RANSAC
Algorithm divides road surface, to extract road surface point cloud data.
Further, the extracting method of curb candidate point includes the road surface point cloud data for obtaining in S2, is swept same
Retouch layer, according between its adjacent scanning element algebraic difference between adjacent gradients feature, difference in height feature and to laser radar distance than feature, will accord with
The scanning element of algebraic difference between adjacent gradients threshold value, difference in height threshold value and distance than threshold value is closed as curb candidate point.
Further, in same scan line, the slope change between point on road surface is more gentle, because
This is by point p in same scan linei=(xi, yi, zi) and its left and right consecutive points between algebraic difference between adjacent gradients as fisrt feature, and position
Slope change between the point on road boundary is larger, and for the road that some rise and fall, road surface is uneven, and phase is only used only
The algebraic difference between adjacent gradients of adjoint point cloud be easy to cause wrong identification as judgment basis.In view of being located at road surface in same scan line
On consecutive points apart from the closely located of laser radar, and distance of the consecutive points apart from laser radar being located on road boundary has
Institute is different, therefore using the distance ratio of consecutive points in same scan line to laser radar as second feature；On the other hand, same
In one scan line, by point pi=(xi, yi, zi) and consecutive points pi+1=(xi+1, yi+1, zi+1) difference in height as third feature.
Algebraic difference between adjacent gradients calculation are as follows:
To laser radar distance than calculation are as follows: Di=disti/disti+1；
The calculation formula of difference in height are as follows: Δ Z=| zi-zi+1|；
Wherein, SiIt is expressed as i-th of scanning element piAlgebraic difference between adjacent gradients, pi=(xi, yi, zi)；I is scanning element in point cloud data
Serial number, x, y, z respectively represent the corresponding x-axis numerical value of each point, y-axis numerical value and z-axis numerical value, i.e. xi, yi, ziRespectively represent each scanning
Point piCorresponding x-axis numerical value, y-axis numerical value and z-axis numerical value；piAdjacent scanning element be respectively pi-1=(xi-1, yi-1, zi-1) and pi+1
=(xi+1, yi+1, zi+1)；threshold1For preset algebraic difference between adjacent gradients threshold value, the threshold2It is preset distance than threshold value,
The threshold3For preset difference in height threshold value, that is, meet following three conditions:
Condition one: Si> threshold1；
Condition two: Di> threshold2；
Condition three: Δ Z > threshold3；
Then by scanning element piAs curb candidate point.
Further, for that can have the pseudo- curb point such as metope, weeds, car in curb candidate point obtained in S3.It examines
The curb dot density for considering same scanning slice is smaller relative to metope and weed density, according to this characteristic, therefore, in step S4
Density Clustering denoising process include using the DBSCAN algorithm based on density to each scanning slice curb candidate point of extraction
Cluster denoising is carried out, if the quantity of core point MinPts is greater than the threshold value of setting, and the radius of neighbourhood in radius of neighbourhood ε
Interior density is less than k scanning element, then using the big scanning element of wherein density and discrete scanning element as noise remove；To each
After scanning slice carries out Density Clustering, the curb candidate point of different scanning slices is polymerized to different classes；K=2 or 3.
Further, described according to having successional feature on road direction, have after cluster is denoised successional
It includes according to Density Clustering as a result, the adjacent class of calculating any two is arrived respectively more than other classes that curb candidate point carries out cluster again
The higher two adjacent classes of similarity are classified as one kind again, and calculate the number for being classified as all kinds of appearance in one kind by string similarity,
The less class of frequency of occurrence is removed, and has the class of intersection to be classified as one kind again after reclassifying, removal does not connect on road direction
Continuous pseudo- curb point, so that continuous curb point is polymerized to one kind.
It further, include: that a kind of point cloud every after clustering again that is to say and be used continuous curb point in the step S6
Weighted least-squares method fitting, finally obtains road edge testing result.
Beneficial effects of the present invention:
(1) compared to most of image processing methods, the present invention detects road environment using laser radar, can
Greatly reduce the environment such as illumination, weather transformation bring curb detection difficult, the present invention has wider applicability；
(2) three-dimensional laser radar that the present invention uses returns to massive point cloud, and data volume is big, can obtain in horizontal direction
360 ° of road environment information, thus its reliability is higher；
(3) present invention carries out multiple features fusion to curb point data to extract curb candidate point, it is ensured that it not will cause missing inspection,
And comprehensively consider curb continuous feature in density and road direction and denoised, so that detection error is small, precision is high.
Detailed description of the invention
Fig. 1 is overall flow schematic diagram of the invention；
Fig. 2 is the experiment scene that embodiment uses in the present invention；
Fig. 3 is the abstract road model and three-dimensional laser radar coordinate system schematic diagram of the present invention；
Fig. 4 is the point cloud chart of a certain layer radar scanning line feature extraction of the present invention；
Fig. 5 is a certain layer radar scanning line density cluster result figure of the present invention；
Fig. 6 is the result figure of detection of the embodiment of the present invention.
Specific embodiment
To make being more clearly understood for the object, technical solutions and advantages of the present invention expression, with reference to the accompanying drawing and specifically
Case study on implementation is described in further details the present invention.
Alternatively, the present embodiment selects Velodyne16 line laser radar as sensor, realizes one kind
Curb detection method based on three-dimensional laser radar, experiment scene is as shown in Fig. 2, specific steps are as follows:
S1, road environment is scanned by three-dimensional laser radar, obtains original point cloud data, and with three-dimensional laser thunder
Up to the three-dimensional system of coordinate for establishing out detection vehicle for origin.
Road environment is scanned by three-dimensional laser radar, the original point cloud data of object is obtained, by laser radar
The point that the laser scanning of transmitting is reflected back to body surface include the three-dimensional coordinate of object, reflected intensity, scanning slice, away from
From information such as, azimuths, the key data of three-dimensional coordinate, distance, scanning slice as laser radar detection lane line is taken, by laser
Radar horizon is mounted on vehicle roof, and using laser radar center as origin o, x, y, z axis passes through origin o and orthogonal foundation is examined
The radar fix system of measuring car, as shown in figure 3, y-axis is parallel to the ground and the headstock of direction direct detection vehicle, x-axis and ground
The right-hand direction of vehicle headstock is directed toward in parallel and direction, and z-axis faces upward perpendicular to ground.
S2, using stochastical sampling consistency algorithm, coarse extraction is carried out to original point cloud data, to extract the point on road surface
Cloud data.
Ground segmentation is carried out to the original point cloud data obtained in S1, data come from Velodyne16 line laser radar, thunder
It is -15 °~15 ° up to vertical scan angle, 16 laser rays is emitted within the scope of vertical scanning, wherein there are 8 scan lines downward
Radiation, 8 scan lines radiate upwards, when detecting curb, need to only consider the scan line of downward radiation, that is, take laser radar downward
Scanning slice；Since radar accuracy is limited, distance is remoter, and single line spacing is bigger, therefore it is each to choose each 20m in vehicle front and back and left and right
10m range is as survey region；Using stochastical sampling consistency (RANdom SAmple Consensus, RANSAC) algorithm pair
Road surface carries out plane fitting, is partitioned into road surface.If the plane expression formula of ground level is ax+by+cz=d, wherein a2+b2+c2=1,
D > 0, (a, b, c) are the direction vector of space line, and d is the distance to plane, this four parameters can determine a plane.
The mode of RANSAC algorithm iteration, picking out from noise-containing ground data includes to count at most in distance threshold
Areal model estimates ground level parameter.Since curb height is generally between 15cm~20cm, setting distance threshold is
0.25m extracts the point cloud data of the road surface of distance fitting or so 0.25m, so as to extract the road surface point comprising curb
Cloud data.
S3, curb data point feature extraction, to extract curb candidate point.
Neighborhood of a point point relationship is respectively scanned using same scan line, constructs a variety of geometrical characteristics, carries out threshold decision, every kind
The threshold value of feature is arranged more loose, filters out and deviates the biggish point of characteristic theory value, guarantees that every kind of feature not will cause missing inspection.
In same scan line, the slope change between point on road surface is more gentle, and is located on road boundary
Slope change between point is larger, therefore by point p in same scan linei=(xi, yi, zi) and its left and right consecutive points between
Algebraic difference between adjacent gradients are as fisrt feature, wherein i is the serial number of each point in point cloud data, and x, y, z respectively represents the corresponding x-axis number of each point
Value, y-axis numerical value and z-axis numerical value, point pi=(xi, yi, zi) algebraic difference between adjacent gradients between consecutive points can indicate are as follows:
Judge algebraic difference between adjacent gradients SiWhether it is greater than given threshold value, that is, judges whether to meet Si> threshold1；For some fluctuatings
Road, road surface is uneven, and the algebraic difference between adjacent gradients that consecutive points cloud is only used only be easy to cause wrong identification as judgment basis.Consider
To in same scan line, the consecutive points on road surface are located at road boundary apart from the closely located of laser radar
On distance of the consecutive points apart from laser radar it is different, therefore by consecutive points in same scan line to the distance of laser radar
Ratio is as second feature, point pi=(xi, yi, zi) and consecutive points pi+1=(xi+1, yi+1, zi+1) to the distance point of laser radar
Not are as follows: distiAnd disti+1, distance ratio can indicate are as follows:
Di=disti/disti+1 (2)
Judge distance ratio DiWhether D is meti> threshold2；In same scan line, by point pi=(xi, yi,
zi) and consecutive points pi+1=(xi+1, yi+1, zi+1) difference in height as third feature, judge whether to meet Δ Z=| zi-zi+1| >
threshold3.The threshold1For preset fisrt feature threshold value, the threshold2For preset second feature threshold
Value, the threshold3For preset third feature threshold value, if meeting above-mentioned judgement, it is determined that the current point pi=
(xi, yi, zi) it is curb candidate point.The point cloud chart of a certain layer radar scanning line is as shown in figure 4, black in figure after feature extraction
The point of color is the data point of a certain scanning slice, and wherein the curb candidate point after feature extraction is indicated with circle.
S4, Density Clustering simultaneously denoise
For that can have the pseudo- curb point such as metope, weeds, car in curb candidate point obtained in S3.It is swept in view of same
The curb dot density for retouching layer is smaller relative to metope and weed density, according to this characteristic, calculates using the DBSCAN based on density
Method carries out cluster denoising to each scanning slice curb candidate point of extraction, if in certain radius of neighbourhood ε, the number of core point MinPts
Amount is greater than the threshold value of setting, and less than 2 or 3 points of density in certain radius of neighbourhood, in the present embodiment, selects less than 3 points；Then
Using the big point of these density and discrete point as noise remove.As shown in figure 5, the point of black is weeds in figure, density is larger,
It is removed as noise.
S5, after cluster denoising will there is successional curb candidate point to cluster again
After carrying out Density Clustering to each scanning slice, the interference of metope and weeds is effectively eliminated, and will be different
Scanning slice curb point is polymerized to different classes；Since curb point has continuity on road direction, according to Density Clustering as a result, meter
The adjacent class of any two is calculated to the cosine similarity of other each classes, the higher two adjacent classes of similarity are classified as one again
Class, and calculate and be classified as the number that a kind of two neighboring class occurs, the less class of removal frequency of occurrence, and a kind of phase will be polymerized to
There is the class of intersection to be classified as one kind in adjacent class, can remove the discontinuous puppet curb point on road direction, and by continuous curb
Point is polymerized to one kind.
In the present embodiment, it is higher similarity can be considered as similarity more than or equal to 99% class, and frequency of occurrence is less than
It is less to be then considered as frequency of occurrence twice or thrice.
S6, fitting curb
The present embodiment is fitted road to a kind of point cloud weighted least-squares method every after clustering again as a preferred method,
Curb, compared to traditional least square method, using the parameter in weighted least-squares method estimation model, by exceptional value influenced compared with
It is small, it can be improved fitting accuracy.It is illustrated in figure 6 the result figure of detection, two straight lines at middle part are curb, accurate detection
The position of curb in the road is gone out.
Based on the above embodiment, it can be seen that the present invention proposes a kind of curb detection method based on three-dimensional laser radar,
Multiple features fusion is carried out to curb point data to extract curb candidate point, has detection accuracy higher, the small advantage of detection error,
And curb can be detected under different illumination conditions, is not illuminated by the light condition influence, stabilization accurately detects road edge.
In order to make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to of the invention real
The technical solution applied in example is clearly and completely described, it is clear that described embodiment is only that present invention a part is implemented
Example, instead of all the embodiments.
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include: ROM, RAM, disk or CD etc..
Embodiment provided above has carried out further detailed description, institute to the object, technical solutions and advantages of the present invention
It should be understood that embodiment provided above is only the preferred embodiment of the present invention, be not intended to limit the invention, it is all
Any modification, equivalent substitution, improvement and etc. made for the present invention, should be included in the present invention within the spirit and principles in the present invention
Protection scope within.
Claims (5)
1. a kind of curb detection method based on three-dimensional laser radar, the described method comprises the following steps:
S1, road environment is scanned by three-dimensional laser radar, obtains original point cloud data, and be with three-dimensional laser radar
Origin establishes out the three-dimensional system of coordinate of detection vehicle；
S2, ground segmentation is carried out using stochastical sampling consistency algorithm to the original point cloud data, extracts the point cloud on road surface
Data；
S3, by the point cloud data on the road surface according to multiple features of curb data point, carry out threshold decision, extract curb time
Reconnaissance；
It is characterized in that, in same scanning slice, according between its adjacent scanning element algebraic difference between adjacent gradients feature, difference in height feature and arrive
Laser radar distance will meet the scanning element of algebraic difference between adjacent gradients threshold value, difference in height threshold value and distance than threshold value as curb than feature
Candidate point；
Algebraic difference between adjacent gradients calculation are as follows:
To laser radar distance than calculation are as follows:
The calculation formula of difference in height are as follows: Δ Z=| zi-zi+1|；
Wherein, SiIt is expressed as i-th of scanning element piAlgebraic difference between adjacent gradients, pi=(xi, yi, zi)；xi, yi, ziRespectively represent each scanning element pi
Corresponding x-axis numerical value, y-axis numerical value and z-axis numerical value；piAdjacent scanning element be respectively pi-1=(xi-1, yi-1, zi-1) and pi+1=
(xi+1, yi+1, zi+1)；threshold1For preset algebraic difference between adjacent gradients threshold value, the threshold2It is preset distance than threshold value, institute
State threshold3For preset difference in height threshold value, that is, meet following three conditions: condition one: Si> threshold1；Condition
Two: Di> threshold2；Condition three: Δ Z > threshold3；Then by scanning element piAs curb candidate point；
S4, according to the density feature of the curb candidate point, the calculation based on density is used to the curb candidate point of each scanning slice
Method carries out cluster denoising；
S5, according on road direction have successional feature, will cluster denoising after have successional curb candidate point again into
Row cluster；
S6, least square method fitting is weighted to the curb candidate point after clustering again；Obtain the testing result of road edge.
2. a kind of curb detection method based on three-dimensional laser radar according to claim 1, which is characterized in that the step
Include the point being reflected back by the laser scanning of three-dimensional laser radar transmitting to body surface in rapid S1, is used to form original point
Cloud data；Wherein original point cloud data includes three-dimensional coordinate, reflected intensity, scanning slice, distance and the azimuth of multiple objects
Deng；Three-dimensional laser radar is horizontally arranged at detection vehicle roof, using three-dimensional laser radar center as origin o, y-axis and ground are flat
The headstock of row and direction direct detection vehicle, x-axis is parallel to the ground and the right-hand direction of vehicle headstock is directed toward in direction, and z-axis is vertical
It faces upward in ground.
3. a kind of curb detection method based on three-dimensional laser radar according to claim 1, which is characterized in that the road
The extracting mode of the point cloud data in face includes the laser radar scanning line for only considering angle downward radiation, and according to the detection of radar
The relationship of precision and distance, research regional scope；Road surface is divided using stochastical sampling consistency algorithm, to extract outlet
Face point cloud data.
4. a kind of curb detection method based on three-dimensional laser radar according to claim 1, which is characterized in that step S4
In Density Clustering denoising process include being carried out using Name-based Routing to each scanning slice curb candidate point of extraction
Cluster denoising, if the quantity of core point MinPts is greater than the threshold value of setting, and in the radius of neighbourhood in radius of neighbourhood ε
Density is less than k scanning element, then using the big scanning element of wherein density and orphan scan point as noise remove；To each scanning
After layer carries out Density Clustering, the curb candidate point of different scanning slices is polymerized to different classes；K=2 or 3.
5. a kind of curb detection method based on three-dimensional laser radar according to claim 4, which is characterized in that described
According to curb there is successional feature on road direction, after cluster denoising will there is successional curb candidate point to gather again
Class includes according to Density Clustering as a result, the cosine similarity that the adjacent class of any two arrives other classes respectively is calculated, by similarity
Higher two adjacent classes are classified as one kind again, calculate the number for being classified as all kinds of appearance in one kind, and removal frequency of occurrence is less
Class, and there is the class of intersection to be classified as one kind, removal discontinuous puppet curb point on road direction, thus will again after reclassifying
Continuous curb candidate point is polymerized to one kind.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910077748.4A CN109738910A (en) | 2019-01-28 | 2019-01-28 | A kind of curb detection method based on three-dimensional laser radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910077748.4A CN109738910A (en) | 2019-01-28 | 2019-01-28 | A kind of curb detection method based on three-dimensional laser radar |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109738910A true CN109738910A (en) | 2019-05-10 |
Family
ID=66366363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910077748.4A Pending CN109738910A (en) | 2019-01-28 | 2019-01-28 | A kind of curb detection method based on three-dimensional laser radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109738910A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110261866A (en) * | 2019-06-12 | 2019-09-20 | 重庆交通大学 | A kind of road width geometry detection method and system based on curb light battle array |
| CN110598541A (en) * | 2019-08-05 | 2019-12-20 | 香港理工大学深圳研究院 | Method and equipment for extracting road edge information |
| CN110736999A (en) * | 2019-10-24 | 2020-01-31 | 北京交通大学 | Railway turnout detection method based on laser radar |
| CN111104908A (en) * | 2019-12-20 | 2020-05-05 | 北京三快在线科技有限公司 | Road edge determination method and device |
| CN111126225A (en) * | 2019-12-17 | 2020-05-08 | 北京易控智驾科技有限公司 | Multiline laser radar ground segmentation method, vehicle and computer readable medium |
| CN111243008A (en) * | 2020-01-19 | 2020-06-05 | 广西师范大学 | Arc data fitting method for high-precision workpiece |
| CN111308499A (en) * | 2020-03-09 | 2020-06-19 | 中振同辂(江苏)机器人有限公司 | Obstacle detection method based on multi-line laser radar |
| CN111401176A (en) * | 2020-03-09 | 2020-07-10 | 中振同辂(江苏)机器人有限公司 | Road edge detection method based on multi-line laser radar |
| CN111580131A (en) * | 2020-04-08 | 2020-08-25 | 西安邮电大学 | Method for identifying vehicles on expressway by three-dimensional laser radar intelligent vehicle |
| CN112368663A (en) * | 2019-11-04 | 2021-02-12 | 深圳市大疆创新科技有限公司 | Terrain prediction method and device for sloping field, radar, unmanned aerial vehicle and operation control method |
| CN112470032A (en) * | 2019-11-04 | 2021-03-09 | 深圳市大疆创新科技有限公司 | Topographic prediction method and device for undulating ground, radar, unmanned aerial vehicle and operation control method |
| CN112488037A (en) * | 2020-12-15 | 2021-03-12 | 上海有个机器人有限公司 | Method for identifying dangerous area in image recognition |
| CN112906813A (en) * | 2021-03-09 | 2021-06-04 | 中南大学 | Flotation condition identification method based on density clustering and capsule neural network |
| CN113138594A (en) * | 2020-01-20 | 2021-07-20 | 北京四维图新科技股份有限公司 | Automatic driving method and device |
| CN113176585A (en) * | 2021-04-14 | 2021-07-27 | 浙江工业大学 | Three-dimensional laser radar-based road surface anomaly detection method |
| CN113702999A (en) * | 2021-07-08 | 2021-11-26 | 中国矿业大学 | Expressway side slope crack detection method based on laser radar |
| CN114170579A (en) * | 2020-08-21 | 2022-03-11 | 广州汽车集团股份有限公司 | Road edge detection method and device and automobile |
| CN115248447A (en) * | 2021-09-29 | 2022-10-28 | 上海仙途智能科技有限公司 | Road edge identification method and system based on laser point cloud |
| CN116047464A (en) * | 2023-01-28 | 2023-05-02 | 武汉理工大学 | Underground mining area gradient detection method and system based on laser radar |
| CN117272086A (en) * | 2023-11-22 | 2023-12-22 | 中国电子科技集团公司第二十九研究所 | Radar signal scanning envelope segmentation method based on DBSCAN |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104636763A (en) * | 2014-12-01 | 2015-05-20 | 北京工业大学 | Road and obstacle detecting method based on remotely piloted vehicles |
| CN104849723A (en) * | 2015-04-14 | 2015-08-19 | 同济大学 | Identification method for simulated lunar surface terrain based on multielement linear array laser radar |
| CN105260737A (en) * | 2015-11-25 | 2016-01-20 | 武汉大学 | Automatic laser scanning data physical plane extraction method with multi-scale characteristics fused |
| US20160026184A1 (en) * | 2014-07-24 | 2016-01-28 | GM Global Technology Operations LLC | Curb detection using lidar with sparse measurements |
| CN105404844A (en) * | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | Road boundary detection method based on multi-line laser radar |
| CN107064955A (en) * | 2017-04-19 | 2017-08-18 | 北京汽车集团有限公司 | barrier clustering method and device |
| CN107169464A (en) * | 2017-05-25 | 2017-09-15 | 中国农业科学院农业资源与农业区划研究所 | A kind of Method for Road Boundary Detection based on laser point cloud |
| CN108519605A (en) * | 2018-04-09 | 2018-09-11 | 重庆邮电大学 | Curb detection method based on laser radar and video camera |
| CN108763287A (en) * | 2018-04-13 | 2018-11-06 | 同济大学 | On a large scale can traffic areas driving map construction method and its unmanned application process |
| CN108828621A (en) * | 2018-04-20 | 2018-11-16 | 武汉理工大学 | Obstacle detection and road surface partitioning algorithm based on three-dimensional laser radar |
| CN108898672A (en) * | 2018-04-27 | 2018-11-27 | 厦门维斯云景信息科技有限公司 | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line |
| CN109017780A (en) * | 2018-04-12 | 2018-12-18 | 深圳市布谷鸟科技有限公司 | A kind of Vehicular intelligent driving control method |
-
2019
- 2019-01-28 CN CN201910077748.4A patent/CN109738910A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160026184A1 (en) * | 2014-07-24 | 2016-01-28 | GM Global Technology Operations LLC | Curb detection using lidar with sparse measurements |
| CN105404844A (en) * | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | Road boundary detection method based on multi-line laser radar |
| CN104636763A (en) * | 2014-12-01 | 2015-05-20 | 北京工业大学 | Road and obstacle detecting method based on remotely piloted vehicles |
| CN104849723A (en) * | 2015-04-14 | 2015-08-19 | 同济大学 | Identification method for simulated lunar surface terrain based on multielement linear array laser radar |
| CN105260737A (en) * | 2015-11-25 | 2016-01-20 | 武汉大学 | Automatic laser scanning data physical plane extraction method with multi-scale characteristics fused |
| CN107064955A (en) * | 2017-04-19 | 2017-08-18 | 北京汽车集团有限公司 | barrier clustering method and device |
| CN107169464A (en) * | 2017-05-25 | 2017-09-15 | 中国农业科学院农业资源与农业区划研究所 | A kind of Method for Road Boundary Detection based on laser point cloud |
| CN108519605A (en) * | 2018-04-09 | 2018-09-11 | 重庆邮电大学 | Curb detection method based on laser radar and video camera |
| CN109017780A (en) * | 2018-04-12 | 2018-12-18 | 深圳市布谷鸟科技有限公司 | A kind of Vehicular intelligent driving control method |
| CN108763287A (en) * | 2018-04-13 | 2018-11-06 | 同济大学 | On a large scale can traffic areas driving map construction method and its unmanned application process |
| CN108828621A (en) * | 2018-04-20 | 2018-11-16 | 武汉理工大学 | Obstacle detection and road surface partitioning algorithm based on three-dimensional laser radar |
| CN108898672A (en) * | 2018-04-27 | 2018-11-27 | 厦门维斯云景信息科技有限公司 | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line |
Non-Patent Citations (5)
| Title |
|---|
| 刘健: "基于三维激光雷达的无人驾驶车辆环境建模关键技术研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 孙朋朋等: "基于3D激光雷达城市道路边界鲁棒检测算法", 《浙江大学学报（工学版）》 * |
| 邹斌等: "基于激光雷达道路可通行区域的检测与提取", 《检测技术与数据处理》 * |
| 魏占营等: "《车载激光测量数据智能后处理技术SWDY深入解析与应用》", 30 November 2018, 武汉大学出版社 * |
| 鲁荣荣等: "一种板型物体混叠场景的快速分割算法", 《光学学报》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110261866A (en) * | 2019-06-12 | 2019-09-20 | 重庆交通大学 | A kind of road width geometry detection method and system based on curb light battle array |
| CN110598541A (en) * | 2019-08-05 | 2019-12-20 | 香港理工大学深圳研究院 | Method and equipment for extracting road edge information |
| CN110736999A (en) * | 2019-10-24 | 2020-01-31 | 北京交通大学 | Railway turnout detection method based on laser radar |
| CN110736999B (en) * | 2019-10-24 | 2021-11-02 | 北京交通大学 | Railway turnout detection method based on laser radar |
| CN112470032A (en) * | 2019-11-04 | 2021-03-09 | 深圳市大疆创新科技有限公司 | Topographic prediction method and device for undulating ground, radar, unmanned aerial vehicle and operation control method |
| CN112368663A (en) * | 2019-11-04 | 2021-02-12 | 深圳市大疆创新科技有限公司 | Terrain prediction method and device for sloping field, radar, unmanned aerial vehicle and operation control method |
| CN111126225B (en) * | 2019-12-17 | 2023-08-04 | 北京易控智驾科技有限公司 | Multi-line laser radar ground segmentation method, vehicle and computer readable medium |
| CN111126225A (en) * | 2019-12-17 | 2020-05-08 | 北京易控智驾科技有限公司 | Multiline laser radar ground segmentation method, vehicle and computer readable medium |
| CN111104908A (en) * | 2019-12-20 | 2020-05-05 | 北京三快在线科技有限公司 | Road edge determination method and device |
| CN111243008A (en) * | 2020-01-19 | 2020-06-05 | 广西师范大学 | Arc data fitting method for high-precision workpiece |
| CN111243008B (en) * | 2020-01-19 | 2022-04-22 | 广西师范大学 | Arc data fitting method for high-precision workpiece |
| CN113138594B (en) * | 2020-01-20 | 2024-04-19 | 北京四维图新科技股份有限公司 | Automatic driving method and device |
| CN113138594A (en) * | 2020-01-20 | 2021-07-20 | 北京四维图新科技股份有限公司 | Automatic driving method and device |
| CN111401176A (en) * | 2020-03-09 | 2020-07-10 | 中振同辂(江苏)机器人有限公司 | Road edge detection method based on multi-line laser radar |
| CN111308499A (en) * | 2020-03-09 | 2020-06-19 | 中振同辂(江苏)机器人有限公司 | Obstacle detection method based on multi-line laser radar |
| CN111580131A (en) * | 2020-04-08 | 2020-08-25 | 西安邮电大学 | Method for identifying vehicles on expressway by three-dimensional laser radar intelligent vehicle |
| CN114170579A (en) * | 2020-08-21 | 2022-03-11 | 广州汽车集团股份有限公司 | Road edge detection method and device and automobile |
| CN112488037A (en) * | 2020-12-15 | 2021-03-12 | 上海有个机器人有限公司 | Method for identifying dangerous area in image recognition |
| CN112488037B (en) * | 2020-12-15 | 2024-05-28 | 上海有个机器人有限公司 | Method for identifying dangerous area in image recognition |
| CN112906813A (en) * | 2021-03-09 | 2021-06-04 | 中南大学 | Flotation condition identification method based on density clustering and capsule neural network |
| CN113176585B (en) * | 2021-04-14 | 2024-03-22 | 浙江工业大学 | Pavement anomaly detection method based on three-dimensional laser radar |
| CN113176585A (en) * | 2021-04-14 | 2021-07-27 | 浙江工业大学 | Three-dimensional laser radar-based road surface anomaly detection method |
| CN113702999A (en) * | 2021-07-08 | 2021-11-26 | 中国矿业大学 | Expressway side slope crack detection method based on laser radar |
| CN115248447A (en) * | 2021-09-29 | 2022-10-28 | 上海仙途智能科技有限公司 | Road edge identification method and system based on laser point cloud |
| CN116047464A (en) * | 2023-01-28 | 2023-05-02 | 武汉理工大学 | Underground mining area gradient detection method and system based on laser radar |
| CN116047464B (en) * | 2023-01-28 | 2023-08-11 | 武汉理工大学 | Underground mining area gradient detection method and system based on laser radar |
| CN117272086B (en) * | 2023-11-22 | 2024-02-13 | 中国电子科技集团公司第二十九研究所 | Radar signal scanning envelope segmentation method based on DBSCAN |
| CN117272086A (en) * | 2023-11-22 | 2023-12-22 | 中国电子科技集团公司第二十九研究所 | Radar signal scanning envelope segmentation method based on DBSCAN |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109738910A (en) | A kind of curb detection method based on three-dimensional laser radar | |
| CN110988912B (en) | Road target and distance detection method, system and device for automatic driving vehicle | |
| CN109858460B (en) | Lane line detection method based on three-dimensional laser radar | |
| CN108509918B (en) | Target detection and tracking method fusing laser point cloud and image | |
| CN109583293B (en) | Aircraft target detection and identification method in satellite-borne SAR image | |
| CN105261020B (en) | A kind of express lane line detecting method | |
| CN105404844B (en) | A kind of Method for Road Boundary Detection based on multi-line laser radar | |
| CN113255481B (en) | Crowd state detection method based on unmanned patrol car | |
| Sportouche et al. | Extraction and three-dimensional reconstruction of isolated buildings in urban scenes from high-resolution optical and SAR spaceborne images | |
| KR101258668B1 (en) | Korea local radar processing system | |
| CN107356933A (en) | A kind of unstructured road detection method based on four line laser radars | |
| CN115761550A (en) | Water surface target detection method based on laser radar point cloud and camera image fusion | |
| CN113281782A (en) | Laser radar snow point filtering method based on unmanned vehicle | |
| CN114325634A (en) | Method for extracting passable area in high-robustness field environment based on laser radar | |
| CN114782729A (en) | Real-time target detection method based on laser radar and vision fusion | |
| Gong et al. | Automated road extraction from LiDAR data based on intensity and aerial photo | |
| Zhang et al. | Nearshore vessel detection based on Scene-mask R-CNN in remote sensing image | |
| CN116573017A (en) | Urban rail train running clearance foreign matter sensing method, system, device and medium | |
| CN113219472B (en) | Ranging system and method | |
| Jiangui et al. | A method for main road extraction from airborne LiDAR data in urban area | |
| Fu et al. | Real-time infrared horizon detection in maritime and land environments based on hyper-laplace filter and convolutional neural network | |
| CN117192513A (en) | Road side laser radar background filtering method in snowy environment | |
| CN116299313A (en) | Laser radar-based intelligent vehicle passable area detection method | |
| CN116165635A (en) | Denoising method for photon cloud data of different beams under daytime condition of multistage filtering algorithm | |
| CN115063698A (en) | Automatic identification and information extraction method and system for slope surface deformation crack |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |
|
| RJ01 | Rejection of invention patent application after publication |